spacy-layout

Why we included this project

spacy-layout is a spaCy plugin that brings PDFs, Word files, and similar input formats into an existing NLP pipeline. It sits on top of Docling and turns a document into a regular spaCy Doc object, with headings, sections, tables, and page layout exposed as labelled spans that map back to the original text. That means the output flows straight into tokenization, named entity recognition, or text classification, and you don't have to handle format conversion yourself. For retrieval teams this is a real advantage: the section and table structure survives chunking, so you can feed a RAG store meaningful pieces with context intact instead of raw page dumps. Table cells come out as pandas DataFrames, and you control how tables render into document text, which keeps extraction code simple.